Cross-Border Claims: 6 Hours to 6 Minutes

How one insurance team cut their international claims processing time by 98% — and caught safety issues they were previously missing.

---

Global Health Partners (GHP) is a mid-sized US health insurance administrator that manages benefit programs for multinational employers. Their member population includes employees based in Germany, Turkey, and Thailand who periodically receive care in the US, and US-based employees who receive care abroad while traveling or on assignment.

In early 2025, GHP processed approximately 200 cross-border claims per month. Each claim arrived with medical codes from the originating country: ICD-10-GM (Germany), ICD-10 WHO-variant (Turkey, Thailand), and pharmacy records referencing local drug brands that US systems don’t recognize.

Two full-time employees spent the majority of their time on these claims. Error rates were climbing. Leadership was looking for a solution.

The Manual Process

Each claim required the same sequence of steps:

1. Identify the coding system. Was this ICD-10-GM, ICD-10 Thai adaptation, or something else? Germany uses the GM variant, maintained by DIMDI/BfArM, with German-specific extensions not present in CM.

2. Look up each code. Manually searching DIMDI’s online ICD-10-GM browser, then cross-referencing to ICD-10-CM. For a typical discharge summary with 8-12 diagnosis codes, this took 15-25 minutes.

3. Handle unmapped codes. ICD-10-GM includes codes specific to German healthcare practice that have no direct CM equivalent. When these appeared, the processor would note “no direct equivalent” and escalate — adding another 20-30 minutes to get clinical review.

4. Resolve drug names. A patient prescribed “Novalgin” in Germany needed to have that flagged as Metamizole — a drug banned by the FDA in the US since 1977 due to agranulocytosis risk. A patient on “ben-u-ron” needed that mapped to acetaminophen (Paracetamol). Without this step, drug interaction checks on US medications failed silently.

5. Document everything. Each mapping decision needed to be logged for audit purposes.

For a straightforward German outpatient claim — 4 diagnosis codes, 2 drugs, standard procedures — the process took approximately 90 minutes end-to-end.

For a complex Turkish hospital discharge with 12 diagnosis codes, comorbidities coded using WHO extension blocks, and 6 medications using Turkish brand names, the process ran 4-6 hours. Sometimes longer.

Where Errors Crept In

The team was diligent. But manual cross-reference lookups against multiple code systems introduced systematic risks:

Code specificity errors. ICD-10-CM is more granular than ICD-10-GM in some categories. K04.01 (reversible pulpitis, CM) and K04.02 (irreversible pulpitis, CM) both map from K04.0 (pulpitis, GM) — but the distinction matters for treatment authorization. Without explicit documentation of which GM code mapped to which CM code, specificity information was routinely lost.

Silent drug misidentification. The most serious category. Turkish claims often referenced drugs by brand names not recognized in US formularies. Without a systematic crosswalk, processors sometimes left drug names untranslated, meaning the drug interaction checker saw unknown entries rather than flagging actual interactions.

Coding system confusion. Thailand uses ICD-10-TM (Thai Modification) with additional codes for tropical diseases and traditional medicine. A processor without specific Thai ICD-10-TM training could mistake a Thai-specific code for a WHO base code and apply incorrect CM equivalents.

Cumulative latency. At 90 minutes per standard claim, the 200-claim monthly volume required approximately 300 hours of staff time — before escalations. Claims waited. Members waited. Clinical decisions downstream were delayed.

TranslateMD Integration

GHP integrated TranslateMD’s API in a two-week sprint. The integration path:

1. Claims intake: Scanned documents fed through TranslateMD’s /translate endpoint with source country specified
2. Structured output: Diagnosis codes mapped from source variant to ICD-10-CM with explicit mapping type (exact/approximate/no-equivalent)
3. Drug crosswalk: Brand names resolved to INN (International Nonproprietary Name) + US generic equivalent + safety flags
4. Human review queue: Claims with safety flags or approximate/no-equivalent mappings routed for clinical review; exact-mapped claims processed automatically

What the API Returns

For a German outpatient claim with the following diagnosis codes:

Input (ICD-10-GM):

K21.0  — Gastroösophageale Refluxkrankheit mit Ösophagitis
J06.9  — Akute Infektion der oberen Atemwege, nicht näher bezeichnet
Z00.0  — Allgemeine ärztliche Untersuchung

Output (ICD-10-CM + mapping metadata):

{
  "diagnoses": [
    {
      "source": { "system": "ICD-10-GM", "code": "K21.0", "description": "Gastroösophageale Refluxkrankheit mit Ösophagitis" },
      "target": { "system": "ICD-10-CM", "code": "K21.0", "description": "Gastro-esophageal reflux disease with esophagitis" },
      "mappingType": "exact",
      "confidence": 1.0
    },
    {
      "source": { "system": "ICD-10-GM", "code": "J06.9", "description": "Akute Infektion der oberen Atemwege, nicht näher bezeichnet" },
      "target": { "system": "ICD-10-CM", "code": "J06.9", "description": "Acute upper respiratory infection, unspecified" },
      "mappingType": "exact",
      "confidence": 1.0
    },
    {
      "source": { "system": "ICD-10-GM", "code": "Z00.0", "description": "Allgemeine ärztliche Untersuchung" },
      "target": { "system": "ICD-10-CM", "code": "Z00.00", "description": "Encounter for general adult medical examination without abnormal findings" },
      "mappingType": "approximate",
      "confidence": 0.85,
      "notes": "GM Z00.0 is less specific than CM Z00.00/Z00.01. CM distinguishes presence of abnormal findings."
    }
  ]
}

For a Turkish prescription referencing Novalgin:

Drug crosswalk output:

{
  "medications": [
    {
      "sourceName": "Novalgin",
      "sourceCountry": "TR",
      "inn": "Metamizole",
      "usGeneric": null,
      "usBrands": null,
      "safetyAlert": {
        "level": "critical",
        "message": "Metamizole (Novalgin/dipyrone) is banned in the US. FDA withdrew approval in 1977 due to risk of agranulocytosis (severe reduction of white blood cells, potentially fatal). Do not substitute — flag for physician review before any US prescription.",
        "action": "HOLD_FOR_CLINICAL_REVIEW"
      }
    }
  ]
}

The Novalgin/Metamizole flag is the kind of catch that manual processing was missing. Metamizole is the most commonly prescribed analgesic in Turkey and Germany. US-based processors, unfamiliar with the drug, were sometimes logging it as “unknown medication” rather than flagging it as a banned substance.

Results After Six Months

GHP tracked outcomes across the first 180 days of TranslateMD integration:

Metric	Before	After	Change
Average processing time (standard claim)	90 min	6 min	−93%
Average processing time (complex claim)	4-6 hours	18-25 min	−93%
Staff time on cross-border claims	~300 hrs/month	~22 hrs/month	−93%
Safety flags generated	0 (process didn’t exist)	34/month average	New capability
Claims requiring clinical escalation	~40% (volume not tracked)	12% (tracked, documented)	Improved
Coding accuracy (audit sample)	Not measured	96.2%	Established baseline

The 34 safety flags per month deserve specific attention. These included:

• Metamizole (Novalgin/Dipyrone) flags: 8-12 per month, primarily from Turkish and German claims. Each flagged for physician review before any US prescription was issued.
• BtM-controlled substances: 4-6 per month. German prescriptions using Betäubungsmittel (BtM) scheduling notation required mapping to DEA scheduling equivalents.
• Approximate mappings requiring review: 18-22 per month. Cases where GM/TH/TR specificity didn’t map exactly to CM — routed for clinical confirmation rather than auto-processed.

The previous process hadn’t surfaced any of these categorically. They either slipped through as unknown or were manually noted without a structured flag process.

The Economics

GHP’s Team plan subscription: $149/month.

Before TranslateMD, the two FTEs dedicated to cross-border claims cost approximately $8,200/month in fully-loaded labor (salary + benefits + overhead allocation). The work wasn’t going away — 200 claims/month was growing as GHP’s international client roster expanded.

After integration:

• Cross-border claims processing: ~22 hours/month of staff time (7% of previous)
• Those staff members reallocated to higher-value work: complex appeals, member services, clinical coordination
• Claims with safety flags receive proper clinical attention instead of silently passing through

The ROI calculation is straightforward: $149/month vs the alternative of adding headcount as volume grows.

What Didn’t Change (By Design)

TranslateMD handles the code translation layer. It doesn’t replace clinical judgment.

Claims with safety flags — Metamizole, approximate diagnosis mappings, unfamiliar procedure codes — still go to a human. What changed is that the human now receives a structured package: the original code, the proposed mapping, the confidence level, and a specific flag with context. The review that used to take 20-30 minutes of lookup and documentation takes 3-4 minutes of evaluation.

The clinical staff doing the review described this as “finally being asked the right question instead of having to figure out what question to ask.”

---

TranslateMD’s API is available on Team ($149/month) and Enterprise plans. The batch processing endpoint handles claim volumes up to 10,000 documents/month on Team; Enterprise supports unlimited volume with SLA guarantees. For insurance and claims processing use cases, contact us at translatemd.io/contact-sales.